基于弱监督的生物医学文献细粒度索引研究

Research on Fine-grained Index of Biomedical Literature Based on Weak Supervision

目前的语义索引难以满足生物医学科研人员的信息检索需求,因而提出一种在概念级别自动优化主题注释的方法,并将其用于阿尔茨海默氏病的MeSH描述符,该描述符对应于代表疾病亚型的6个不同概念。结果表明,所提出的基于弱监督的索引方案可以改善文字字符串匹配的性能。此外,随着MeSH词库条目信息逐渐完善,所提方案可以提供更加精确的搜索,还可以将主题注释与其他语义信息集成在一起,有助于维护主题注释的一致性。

The current semantic index cannot meet the information needs of researchers.This paper proposes a method for automatically optimizing topic annotations at the concept level and uses it for the MeSH descriptor of Alzheimer’s disease,which corresponds to six different concepts representing disease subtypes.The researching results show that the indexing scheme based on weak supervision proposed in this paper can improve the performance of text string matching.In addition,with the gradual improvement of MeSH thesaurus entry information,the solution proposed in this article can provide more accurate search,and can also integrate topic annotations with other semantic information,which helps maintain the consistency of topic annotations.